Re: [PATCH v2 11/18] hw/block/nvme: add remaining mandatory controller parameters

[Klaus Jensen]

On Jul  3 13:02, Philippe Mathieu-Daudé wrote:
> On 7/3/20 12:10 PM, Klaus Jensen wrote:
> > On Jul  3 11:21, Philippe Mathieu-Daudé wrote:
> >> On 7/3/20 10:46 AM, Klaus Jensen wrote:
> >>> On Jul  3 10:31, Philippe Mathieu-Daudé wrote:
> >>>> On 7/3/20 8:34 AM, Klaus Jensen wrote:
> >>>>> From: Klaus Jensen <k.jensen@samsung.com>
> >>>>>
> >>>>> Add support for any remaining mandatory controller operating parameters
> >>>>> (features).
> >>>>>
> >>>>> Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
> >>>>> Reviewed-by: Dmitry Fomichev <dmitry.fomichev@wdc.com>
> >>>>> ---
> >>>>>  hw/block/nvme.c       | 39 +++++++++++++++++++++++++++++++++------
> >>>>>  hw/block/nvme.h       | 18 ++++++++++++++++++
> >>>>>  hw/block/trace-events |  2 ++
> >>>>>  include/block/nvme.h  |  7 +++++++
> >>>>>  4 files changed, 60 insertions(+), 6 deletions(-)
> >>>>>
> >>>>> diff --git a/hw/block/nvme.c b/hw/block/nvme.c
> >>>>> index ba523f6768bf..affb9a967534 100644
> >>>>> --- a/hw/block/nvme.c
> >>>>> +++ b/hw/block/nvme.c
> >>>>> @@ -1056,8 +1056,16 @@ static uint16_t nvme_get_feature(NvmeCtrl *n, 
> >>>>> NvmeCmd *cmd, NvmeRequest *req)
> >>>>>      uint32_t dw10 = le32_to_cpu(cmd->cdw10);
> >>>>>      uint32_t dw11 = le32_to_cpu(cmd->cdw11);
> >>>>>      uint32_t result;
> >>>>> +    uint8_t fid = NVME_GETSETFEAT_FID(dw10);
> >>>>> +    uint16_t iv;
> >>>>>  
> >>>>> -    switch (dw10) {
> >>>>> +    trace_pci_nvme_getfeat(nvme_cid(req), fid, dw11);
> >>>>> +
> >>>>> +    if (!nvme_feature_support[fid]) {
> >>>>> +        return NVME_INVALID_FIELD | NVME_DNR;
> >>>>> +    }
> >>>>> +
> >>>>> +    switch (fid) {
> >>>>>      case NVME_TEMPERATURE_THRESHOLD:
> >>>>>          result = 0;
> >>>>>  
> >>>>> @@ -1088,14 +1096,27 @@ static uint16_t nvme_get_feature(NvmeCtrl *n, 
> >>>>> NvmeCmd *cmd, NvmeRequest *req)
> >>>>>                               ((n->params.max_ioqpairs - 1) << 16));
> >>>>>          trace_pci_nvme_getfeat_numq(result);
> >>>>>          break;
> >>>>> +    case NVME_INTERRUPT_VECTOR_CONF:
> >>>>> +        iv = dw11 & 0xffff;
> >>>>> +        if (iv >= n->params.max_ioqpairs + 1) {
> >>>>> +            return NVME_INVALID_FIELD | NVME_DNR;
> >>>>> +        }
> >>>>> +
> >>>>> +        result = iv;
> >>>>> +        if (iv == n->admin_cq.vector) {
> >>>>> +            result |= NVME_INTVC_NOCOALESCING;
> >>>>> +        }
> >>>>> +
> >>>>> +        result = cpu_to_le32(result);
> >>>>> +        break;
> >>>>>      case NVME_ASYNCHRONOUS_EVENT_CONF:
> >>>>>          result = cpu_to_le32(n->features.async_config);
> >>>>>          break;
> >>>>>      case NVME_TIMESTAMP:
> >>>>>          return nvme_get_feature_timestamp(n, cmd);
> >>>>>      default:
> >>>>> -        trace_pci_nvme_err_invalid_getfeat(dw10);
> >>>>> -        return NVME_INVALID_FIELD | NVME_DNR;
> >>>>> +        result = cpu_to_le32(nvme_feature_default[fid]);
> >>>>
> >>>> So here we expect uninitialized fid entries to return 0, right?
> >>>>
> >>>
> >>> Yes, if defaults are not 0 (like NVME_ARBITRATION), it is explicitly set.
> >>>
> >>>>> +        break;
> >>>>>      }
> >>>>>  
> >>>>>      req->cqe.result = result;
> >>>>> @@ -1124,8 +1145,15 @@ static uint16_t nvme_set_feature(NvmeCtrl *n, 
> >>>>> NvmeCmd *cmd, NvmeRequest *req)
> >>>>>  {
> >>>>>      uint32_t dw10 = le32_to_cpu(cmd->cdw10);
> >>>>>      uint32_t dw11 = le32_to_cpu(cmd->cdw11);
> >>>>> +    uint8_t fid = NVME_GETSETFEAT_FID(dw10);
> >>>>>  
> >>>>> -    switch (dw10) {
> >>>>> +    trace_pci_nvme_setfeat(nvme_cid(req), fid, dw11);
> >>>>> +
> >>>>> +    if (!nvme_feature_support[fid]) {
> >>>>> +        return NVME_INVALID_FIELD | NVME_DNR;
> >>>>> +    }
> >>>>> +
> >>>>> +    switch (fid) {
> >>>>>      case NVME_TEMPERATURE_THRESHOLD:
> >>>>>          if (NVME_TEMP_TMPSEL(dw11) != NVME_TEMP_TMPSEL_COMPOSITE) {
> >>>>>              break;
> >>>>> @@ -1172,8 +1200,7 @@ static uint16_t nvme_set_feature(NvmeCtrl *n, 
> >>>>> NvmeCmd *cmd, NvmeRequest *req)
> >>>>>      case NVME_TIMESTAMP:
> >>>>>          return nvme_set_feature_timestamp(n, cmd);
> >>>>>      default:
> >>>>> -        trace_pci_nvme_err_invalid_setfeat(dw10);
> >>>>> -        return NVME_INVALID_FIELD | NVME_DNR;
> >>>>> +        return NVME_FEAT_NOT_CHANGEABLE | NVME_DNR;
> >>>>>      }
> >>>>>      return NVME_SUCCESS;
> >>>>>  }
> >>>>> diff --git a/hw/block/nvme.h b/hw/block/nvme.h
> >>>>> index f8940435f9ef..8ad1e3c89cee 100644
> >>>>> --- a/hw/block/nvme.h
> >>>>> +++ b/hw/block/nvme.h
> >>>>> @@ -87,6 +87,24 @@ typedef struct NvmeFeatureVal {
> >>>>>      uint32_t    async_config;
> >>>>>  } NvmeFeatureVal;
> >>
> >> What do you think about adding:
> >>
> >> --- a/include/block/nvme.h
> >> +++ b/include/block/nvme.h
> >> @@ -804,7 +804,8 @@ enum NvmeFeatureIds {
> >>      NVME_WRITE_ATOMICITY            = 0xa,
> >>      NVME_ASYNCHRONOUS_EVENT_CONF    = 0xb,
> >>      NVME_TIMESTAMP                  = 0xe,
> >> -    NVME_SOFTWARE_PROGRESS_MARKER   = 0x80
> >> +    NVME_SOFTWARE_PROGRESS_MARKER   = 0x80,
> >> +    NVME_FEATURE_ID_COUNT           = 0x100
> >>  };
> >>
> > 
> > I can definitely go with that. I added it as NVME_FID_MAX.
> 
> Good name.
> 
> > 
> >>>>>  
> >>>>> +static const uint32_t nvme_feature_default[0x100] = {
> >>
> >> Why uint32_t and not uint16_t?
> >>
> > 
> > The feature values are by definition 32 bits wide, so even though the
> > defaults here do not require it, I think uin32_t should be used.
> 
> Can you prepend a new patch making nvme_get_feature() return uin32_t
> instead of uint16_t? That would be clearer.
> 

Ah. Now I see where the confusion comes from ;)

nvme_get_feature() does not return the value of the feature. It returns
a uint16_t with a value from the NvmeStatusCodes enum (which really
should be a typedef used all over the place.